Author Introduction
Kia ora, I’m Andrew McPherson. In my work helping B2B teams win visibility across AI answer engines, the single biggest lift I see comes from structured product and service feeds. If your offerings aren’t machine-readable, AI models simply can’t cite them. Here’s how to fix that.
Outline
- What product and service feed optimisation means
- Why AI systems prefer structured feeds over prose
- How AI models retrieve and compare feed data
- Four-layer schema optimisation approach explained
- Worked Product, Service and SoftwareApplication examples
- Validation tools and common deployment errors
- CiteCompass perspective on feed prioritisation
- Recent schema and AI platform changes
Key Takeaways
- Structured feeds beat prose for AI commercial queries
- Use Product, Service or SoftwareApplication schema types
- Complete recommended properties lift AI retrieval priority
- Offer schema enables accurate pricing and availability answers
- Validate every schema with Rich Results Test
- Start with top ten revenue-generating offerings first
- Nested entity references build usable knowledge graphs
- Schema completeness outperforms keyword density for citation
What Is Product and Service Feed Optimisation?
Product and service feed optimisation is the practice of structuring data about commercial offerings using standardised schema markup so AI systems can accurately discover, understand and cite your products or services. It transforms unstructured web content (product pages, service descriptions, case studies) into machine-readable entities that AI models can parse during retrieval-augmented generation (RAG) processes.
Unlike generic web content optimisation, feed optimisation focuses on three core schema types defined by Schema.org: Product (for physical goods and B2B equipment), Service (for professional services and consulting), and SoftwareApplication (for SaaS products and enterprise software). Each type includes properties for pricing, availability, specifications and provider information that AI systems use when answering comparative queries, feature lookups or purchasing guidance questions.
Optimisation occurs at the structured data layer. You add JSON-LD markup to existing product and service pages to create a parallel feed AI systems can consume without parsing prose. This dual-layer approach (human-readable page plus machine-readable feed) is what Microsoft describes as the “feeds and APIs” data surface in its AI visibility framework.
Why Product and Service Feeds Matter for AI Systems
AI systems increasingly answer commercial intent queries by retrieving structured data rather than summarising marketing copy. When a user asks ChatGPT “What CRM systems support custom workflows?” or Google AI Overview answers “Which industrial sensors measure temperature and pressure?”, the model retrieves and compares Product or Service schema from multiple sources.
Three operational reasons explain the preference for structured feeds over unstructured content.
First, structured data enables entity disambiguation. AI models need to distinguish between “Salesforce the company” and “Salesforce Sales Cloud the product”, or between generic “consulting services” and specific “IT security consulting”. Schema markup provides explicit @type declarations, unique identifiers and hierarchical relationships that prevent conflation.
Second, structured feeds support feature-level comparisons. When AI systems generate comparison tables or answer specification queries, they extract data from standardised properties like featureList, applicationCategory and offers. Prose descriptions require sentence parsing and entity extraction, which introduces errors. Structured data provides pre-parsed facts.
Third, pricing and availability information requires precision. AI systems cannot reliably extract tiered pricing, volume discounts or lead times from marketing language. Properties like priceSpecification, eligibleQuantity and deliveryLeadTime give AI models the specificity needed to answer transactional queries accurately. Google Search Central guidance emphasises complete Offer and rating properties for product information enhancement in AI-powered results.
For B2B companies, this matters more than consumer e-commerce because commercial queries involve complex decision criteria. Business buyers ask multi-faceted questions (“What project management software supports Gantt charts, integrates with Jira and costs under $50 per user?”). AI systems answer by filtering structured Product or Service data, not by reading feature blog posts.
How AI Systems Use Product and Service Data
AI systems access product and service data through three distinct mechanisms, each with different schema requirements.
Direct Retrieval During Query Processing
When an AI model receives a query, it searches indexed structured data for entities matching query parameters. For product queries, it filters by category, brand, offers.price and additionalProperty (technical specifications). For service queries, it filters by serviceType, areaServed and provider. Each schema property behaves like an indexed database column.
Knowledge Graph Construction
AI systems build temporary knowledge graphs by linking related entities through schema references. A Product references its manufacturer (Organization), reviews (AggregateRating) and sellers (Offer). The model traverses Product > manufacturer > Organization paths to identify companies, then aggregates results. Graph traversal only works if schema uses consistent @id references and declared entity types.
Feature Extraction for Comparison
AI systems parse featureList arrays, description fields and additionalProperty objects to extract capabilities. For SaaS products this includes integrations, API availability and security certifications. For professional services this includes methodologies, certifications and delivery models. Extracted features populate comparison matrices and capability answers.
Schema.org documentation specifies required and recommended properties for each schema type. For Product schema, required properties include name, description and offers. Recommended properties include brand, sku, image, aggregateRating and review. For Service schema, required properties include name and serviceType, with provider, areaServed, offers and termsOfService recommended. AI systems give higher retrieval priority to entities with complete recommended properties, because incomplete schemas provide insufficient context for accurate citation.
How to Optimise Product and Service Feeds
Optimisation follows a four-layer approach: entity completeness, property richness, nested relationships and validation. Each layer builds on the previous to maximise AI retrieval accuracy.
Layer 1: Entity Completeness
Start with the base schema type that matches your offering. Use Product for physical goods, manufactured equipment or hardware. Use Service for professional services, consulting, training or support contracts. Use SoftwareApplication for SaaS products, enterprise applications or licensed software. Declare the @type explicitly and include a globally unique @id following a consistent pattern across your domain.
Every entity must include: name (official product or service name), description (a 150-250 word explanation of what it does and who it serves), image (a high-resolution product image or service visualisation), and url (the canonical page URL). These four properties form the minimum viable entity that AI systems will index.
Layer 2: Property Richness
Add recommended properties that provide filterable attributes. For Product schema include brand, manufacturer, sku, gtin or mpn, and category. For Service schema include serviceType, provider, areaServed and availableChannel.
Technical specifications belong in additionalProperty, which accepts an array of PropertyValue objects. Each PropertyValue includes a name (specification label) and value (specification data). B2B product examples include operating temperature range, power requirements, compliance certifications, dimensions and weight. Service examples include response time SLAs, delivery timelines, minimum engagement duration and client prerequisites.
Layer 3: Nested Relationships
Connect your Product or Service entity to related entities using proper references. The most critical nested entity is Offer, which contains pricing and availability data. Create an Offer object for each pricing tier, region or customer segment. Include price, priceCurrency (ISO 4217), priceSpecification, availability, and validFrom/validThrough dates.
For products with customer reviews, add an AggregateRating object with ratingValue, bestRating, worstRating and ratingCount. AI systems use ratings as quality signals when ranking retrieved products. For services, include hasOfferCatalog if you offer multiple related services, linking to an OfferCatalog entity that lists service packages.
Layer 4: Validation and Deployment
Test schema markup with Google’s Rich Results Test and Schema.org’s validator. These tools check JSON-LD syntax, required properties and type correctness. Fix errors before deployment. Common errors include missing required properties, incorrect data types (strings instead of numbers for prices), invalid ISO codes for currencies or availability, and broken @id references.
Worked Example: Product Schema
A complete Product schema for a B2B industrial sensor:
{
“@context”: “https://schema.org”,
“@type”: “Product”,
“@id”: “https://example.com/products/temp-sensor-tx400#product”,
“name”: “TX400 Industrial Temperature Sensor”,
“description”: “High-precision temperature sensor for industrial automation systems. Measures -50°C to 300°C with ±0.5°C accuracy. IP67 rated housing with 4-20mA output.”,
“brand”: { “@type”: “Organization”, “name”: “Example Sensors Inc.”, “@id”: “https://example.com/#organization” },
“manufacturer”: { “@id”: “https://example.com/#organization” },
“sku”: “TX400-001”,
“mpn”: “TX400”,
“image”: “https://example.com/images/tx400-sensor.jpg”,
“category”: “Industrial Sensors”,
“additionalProperty”: [
{ “@type”: “PropertyValue”, “name”: “Temperature Range”, “value”: “-50°C to 300°C” },
{ “@type”: “PropertyValue”, “name”: “Accuracy”, “value”: “±0.5°C” },
{ “@type”: “PropertyValue”, “name”: “Output Signal”, “value”: “4-20mA” },
{ “@type”: “PropertyValue”, “name”: “IP Rating”, “value”: “IP67” }
],
“offers”: {
“@type”: “Offer”,
“price”: “285.00”,
“priceCurrency”: “USD”,
“availability”: “https://schema.org/InStock”,
“priceSpecification”: {
“@type”: “PriceSpecification”,
“minPrice”: “285.00”,
“priceCurrency”: “USD”,
“eligibleQuantity”: { “@type”: “QuantitativeValue”, “minValue”: 1, “maxValue”: 9 }
},
“seller”: { “@id”: “https://example.com/#organization” }
},
“aggregateRating”: { “@type”: “AggregateRating”, “ratingValue”: “4.7”, “bestRating”: “5”, “worstRating”: “1”, “ratingCount”: “23” }
}
Worked Example: Service Schema
A complete Service schema for professional consulting:
{
“@context”: “https://schema.org”,
“@type”: “Service”,
“@id”: “https://example.com/services/cybersecurity-consulting#service”,
“name”: “Cybersecurity Consulting Services”,
“description”: “Comprehensive cybersecurity consulting for mid-market B2B companies. Services include security audits, penetration testing, compliance assessment (SOC 2, ISO 27001) and incident response planning.”,
“serviceType”: “Cybersecurity Consulting”,
“provider”: { “@type”: “Organization”, “name”: “Example Security Partners”, “@id”: “https://example.com/#organization” },
“areaServed”: { “@type”: “Place”, “name”: “United States” },
“availableChannel”: { “@type”: “ServiceChannel”, “serviceType”: “Remote and On-site” },
“offers”: {
“@type”: “Offer”,
“price”: “15000.00”,
“priceCurrency”: “USD”,
“priceSpecification”: {
“@type”: “PriceSpecification”,
“price”: “15000.00”,
“priceCurrency”: “USD”,
“name”: “Base Engagement Fee”,
“description”: “Minimum engagement covers initial security assessment and report”
}
},
“termsOfService”: “https://example.com/services/terms”,
“additionalProperty”: [
{ “@type”: “PropertyValue”, “name”: “Engagement Duration”, “value”: “4-12 weeks” },
{ “@type”: “PropertyValue”, “name”: “Certifications”, “value”: “CISSP, CEH, CISM” }
]
}
Worked Example: SoftwareApplication Schema
A SoftwareApplication schema for a SaaS product:
{
“@context”: “https://schema.org”,
“@type”: “SoftwareApplication”,
“@id”: “https://example.com/products/project-suite#software”,
“name”: “Project Suite Pro”,
“description”: “Cloud-based project management software for distributed teams. Features include Gantt charts, resource allocation, time tracking, budget management and integrations with Slack, Jira and GitHub.”,
“applicationCategory”: “ProjectManagementApplication”,
“operatingSystem”: “Web-based (Chrome, Firefox, Safari, Edge)”,
“offers”: {
“@type”: “Offer”,
“price”: “49.00”,
“priceCurrency”: “USD”,
“priceSpecification”: {
“@type”: “UnitPriceSpecification”,
“price”: “49.00”,
“priceCurrency”: “USD”,
“referenceQuantity”: { “@type”: “QuantitativeValue”, “value”: “1”, “unitText”: “user per month” }
}
},
“featureList”: [
“Gantt charts and timeline views”,
“Resource allocation and workload management”,
“Time tracking with billable hours”,
“Budget tracking and forecasting”,
“Slack integration”,
“Jira integration”,
“GitHub integration”,
“REST API access”
],
“softwareVersion”: “3.2”,
“aggregateRating”: { “@type”: “AggregateRating”, “ratingValue”: “4.5”, “bestRating”: “5”, “ratingCount”: “142” }
}
CiteCompass Perspective
CiteCompass treats product and service feed optimisation as part of the broader AI Data Surfaces strategy. Structured feeds are the second data surface (feeds and APIs) that AI systems access independently from crawled web content. Many B2B companies already have product or service pages but lack the structured data layer that makes those offerings discoverable to AI models.
Our approach prioritises schema completeness over keyword density. AI systems do not rank structured data by keyword frequency. They filter and retrieve based on declared properties and entity relationships. A Product schema with complete technical specifications, tiered pricing and manufacturer references will outperform a keyword-stuffed description with minimal structured data.
We recommend starting with your top 10 revenue-generating products or services. Implement complete schema markup for those offerings first, validate using Google’s Rich Results Test, and monitor AI citation frequency. Expand to additional products once you establish a reliable deployment workflow. For large catalogues, consider automated schema generation from product databases or CMS structured fields, but always validate output quality before deployment. Our AI Visibility Suite measures citation lift at the entity level so you can see which feeds are pulling their weight.
What Changed Recently
- 2025-01: Google Search Central updated Product schema documentation emphasising aggregateRating and review properties for AI Overview inclusion.
- 2024-12: Schema.org version 16.0 added new properties for subscription-based pricing models under Offer.priceSpecification, benefiting SaaS products.
- 2024-Q3: Microsoft Bing AI began indexing Service schema, expanding AI citation opportunities beyond Product markup.
Related Topics
Explore related concepts in the Technical Implementation pillar:
Learn about AI Data Surfaces in the Core Frameworks pillar, or return to the CiteCompass Knowledge Hub.
References
- Schema.org (2024). Product Schema Documentation. https://schema.org/Product – official specification defining Product structured data properties used by AI systems.
- Google Search Central (2024). Use structured data to enhance product information in Search. https://developers.google.com/search/docs/appearance/structured-data/product – guidelines emphasising aggregateRating and review for AI Overview inclusion.
- Schema.org (2024). Service Schema Documentation. https://schema.org/Service – official specification for Service structured data enabling AI citation of professional services.